The Anthrope American Artificial Intelligence Start-Up (AI) has agreed to pay at least $ 1.5 billion to a compensation fund for authors, beneficiaries and editors who continued the company for illegally downloading millions of pounds, according to a judicial document published Friday.
This friendly agreement, with a colossal amount, constitutes a surprising step in the debate on the use of data to develop and cause large models of generative AI.
“This historical agreement is the highest for a copyright case,” said the AFP lawyer for rights holders, Justin Nelson, of the firm Susman Godfrey.
$ 1.5 billion
At the end of June, the Seized Californian judge had estimated that the fact of feeding the generative AI software with works in theory protected by copyright did not constitute a crime.
He had only preserved against Anthrope the download and storage of online pirate bookstores books, recognizing the Californian company guilty of not having bought these works.
“We do not agree with the opinion of the Court according to which we can separate the discharge of a job of its use,” said an Anthrope spokesman. But “we believe that this agreement will allow us to focus on our essential mission, instead of a long litigation,” he added.
The amount on which the parties have been granted will be at least $ 1.5 billion and could increase if the final list of the books in question, which has not yet been arrested, exceeded 500,000, in which case Anthrope would pay $ 3,000 more for work.
The agreement must still be approved by Judge William Alsup. A hearing is scheduled for Monday in the Federal Court of San Francisco.
The notion of “fair use” or fair use
This transaction allows anthropic to avoid an essay, which would begin in early December to determine the amount of damage. It is likely that the new company condemned to pay a sum of what is decided with the holders of the law, to the point of endangering its own existence.
Many other files are still ongoing before the US courts, initiated by writers, musicians or press editors for the unauthorized use of their production.
On Friday, two writers launched an appeal, who want collective, against Apple, accusing the Silicon Valley giant of having used works contained in pirate libraries to train the models of AI integrated into their devices.
The majority of the main actors in the generative AI trust the legal concept of fair use (“fair use” in English), which probably limit the application of intellectual property right.
“Creative content theft”
In his June decision, Judge Alsup had considered that when drawing his AI models, called Claude, with thousands of books, Anthrope was part of the fair use.
“The legal principle according to which the development of AI in protected works is of fair use remains intact,” said Anthrope spokesman. This agreement “only resolves a dispute on how certain documents have been obtained,” he continued.
In June, in another case of this type, with respect to goal, before the same jurisdiction, another federal magistrate had given the giant of social networks, but by explaining that the plaintiffs could have raised admissible arguments.
Content creators, either music, books or articles, are concerned with seeing the market value of their labor collapse with the appearance of generative interfaces of AI.
“This is just the beginning,” the association wrote in X, “but it is surprising and historical.”
Source: BFM TV
